Accession Number:

ADA581312

Title:

UGent Participation in the Microblog Track 2012

Descriptive Note:

Conference paper

Corporate Author:

GHENT UNIV (BELGIUM)

Report Date:

2012-11-01

Pagination or Media Count:

6.0

Abstract:

In this paper, we describe the search system, developed at Ghent University for the TREC 2012 Microblog Track in order to rank Twitter messages or tweet from a fixed corpus in response to a number of search requests. Our system ranks the tweets based on a Logistic Regression classifier trained with data from the Microblog Track 2011. The features used for training the classifier include local tweets features, but also, query expansion and tweet expansion features, based on external Web data, which appear to significantly improve results.

Subject Categories:

  • Information Science

Distribution Statement:

APPROVED FOR PUBLIC RELEASE